Refactor softmax templates to use outer dims #844
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
facebook-github-bot
added
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D47732859 |
int3
added a commit
to int3/AITemplate
that referenced
this pull request
Jul 25, 2023
Summary: Pull Request resolved: facebookincubator#844 Previously, the softmax templates assumed that reduction would always be done over the last dim, so the only parameter passed to the templates was the rank of the tensor. To set the stage for generalizing softmax, we pass the reduction dim instead. The output is functionally identical, though the codegen changes slightly in the case where all the inner dimensions are 1: we now pass only the outer dimensions to the function call, dropping the redundant inner dimension parameters. For the `tail_shapes_all_1_bf16` softmax test case, we have Before: ``` softmax_0( X, Y, &input_batch, &X_dim_1, &X_dim_2, stream ); ``` After: ``` softmax_0( X, Y, &input_batch, stream ); ``` Differential Revision: D47732859 fbshipit-source-id: 604ef9901036b913d5f8bc0ce779fb9c3f4b6aa2
|
This pull request was exported from Phabricator. Differential Revision: D47732859 |
Summary: Very minor cleanups I did while familiarizing myself with the code Aside from whitespace changes I also removed a few unnecessary automatic variables Differential Revision: D47732846 fbshipit-source-id: 88635358cf795f21ce1c582dea85185db6c5ac07
Summary: Pull Request resolved: facebookincubator#844 Previously, the softmax templates assumed that reduction would always be done over the last dim, so the only parameter passed to the templates was the rank of the tensor. To set the stage for generalizing softmax, we pass the reduction dim instead. The output is functionally identical, though the codegen changes slightly in the case where all the inner dimensions are 1: we now pass only the outer dimensions to the function call, dropping the redundant inner dimension parameters. For the `tail_shapes_all_1_bf16` softmax test case, we have Before: ``` softmax_0( X, Y, &input_batch, &X_dim_1, &X_dim_2, stream ); ``` After: ``` softmax_0( X, Y, &input_batch, stream ); ``` Reviewed By: aakhundov Differential Revision: D47732859 fbshipit-source-id: 9c081a21338ba5d6cb6386b06b113171611746e4
|
This pull request was exported from Phabricator. Differential Revision: D47732859 |
|
This pull request has been merged in ecf6037. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary:
Previously, the softmax templates assumed that reduction would always be done over the last dim, so the only parameter passed to the templates was the rank of the tensor. To set the stage for generalizing softmax, we pass the reduction dim instead.
The output is functionally identical, though the codegen changes slightly in the case where all the inner dimensions are 1: we now pass only the outer dimensions to the function call, dropping the redundant inner dimension parameters.
For the
tail_shapes_all_1_bf16softmax test case, we haveBefore:
After:
Differential Revision: D47732859